Edge Sign Classification

Hi, I want to perform binary edge sign classification task. Edges in the graph are of the same type but with positive/negative signs. I couldn’t find a clear way to implement it using DGL. Would love to hear some advice.

I’m thinking about running unsupervised GAT to generate node embeddings then train a normal classifier on concatenation of source||target. The unsupervised version is not mentioned in the paper so I’m following the this Link Prediction tutorial. Are there more tutorials on unsupervised nodes encoders?


1 Like

The link prediction section here might be helpful.

1 Like